三星索引概念：
第一颗星：与查询相关的索引行是相邻的，也就是where后面的等值谓词，可以匹配索引列顺序
第二颗星：索引行的顺序与查询语句需求一致，也就是order by 中的排序和索引顺序是否一致
第三颗星：索引行包含查询语句中所有的列
参考：《数据库索引设计与优化》
一、建立测试数据
SQL> create table id.cust(cno int,lname varchar(10),fname varchar(10),sex int,weight int,city varchar(10));
SQL> truncate table id.cust;
SQL> declare
i number;
begin
for i in 1..1000 loop
insert into id.cust
values(i, DBMS_RANDOM.STRING('u', 4), DBMS_RANDOM.STRING('u', 4),ABS(MOD(DBMS_RANDOM.RANDOM,2)),ABS(MOD(DBMS_RANDOM.RANDOM,150)),DBMS_RANDOM.STRING('u', 6));
commit;
end loop;
end;
二、三星索引测试
1、满足三星索引
--建立索引(三星)
SQL> create index id.idx_cust on id.cust(city,lname,fname,cno);
对于下面的sql，建立如上就是三星索引：
第一颗星：所有等值谓词的列，把这些做组合索引的开头的列，如下sql的where条件的city,lname开头，顺序随意。
第二颗星：order by的字段加入组合索引中 fname字段(索引自动排序)
第三颗星：将查询中剩余的列加入组合索引中 select中的cno字段(fname之前已加入过)
SQL> EXPLAIN PLAN FOR
2 select cno,fname
3 from id.cust
4 where lname = :1
5 and city = :2
6 order by fname;
Explained.
这样建立索引后，执行计划最优，如下：
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------
Plan hash value: 3860859945
-----------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-----------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 34 | 2 (0)| 00:00:01 |
|* 1 | INDEX RANGE SCAN| IDX_CUST | 1 | 34 | 2 (0)| 00:00:01 |
-----------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------
1 - access("CITY"=:2 AND "LNAME"=:1)
Note
-----
- dynamic statistics used: dynamic sampling (level=2)
17 rows selected.
2、不满足三星索引
如果在where条件中，不满足第一颗星 等值谓词，那之前组合索引的顺序要更改
SQL> EXPLAIN PLAN FOR
2 select cno,fname
3 from id.cust
4 where lname between :lname1 and :lname2
5 and city = :2
6 order by fname;
Explained.
执行计划有filter过滤，还有order by排序操作， 不满足第一颗星和第二颗星
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------
Plan hash value: 1720078986
-------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 34 | 3 (34)| 00:00:01 |
| 1 | SORT ORDER BY | | 1 | 34 | 3 (34)| 00:00:01 |
|* 2 | FILTER | | | | | |
|* 3 | INDEX RANGE SCAN| IDX_CUST | 1 | 34 | 2 (0)| 00:00:01 |
-------------------------------------------------------------------------------
PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
2 - filter(:LNAME2>=:LNAME1)
3 - access("CITY"=:2 AND "LNAME">=:LNAME1 AND "LNAME"<=:LNAME2)
Note
-----
- dynamic statistics used: dynamic sampling (level=2)
20 rows selected.
修改一下组合索引，让order by的列fname放在between谓词列lname之前，才能满足第二颗星，避免排序操作
SQL> drop index id.idx_cust ;
Index dropped.
SQL> create index id.idx_cust on id.cust(city,fname,lname,cno);
Index created.
SQL> EXPLAIN PLAN FOR
2 select cno,fname
3 from id.cust
4 where lname between :lname1 and :lname2
5 and city = :2
6 order by fname;
Explained.
查看执行计划，没有了order by排序操作(索引自动排序)
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------
Plan hash value: 1083764640
------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 34 | 2 (0)| 00:00:01 |
|* 1 | FILTER | | | | | |
|* 2 | INDEX RANGE SCAN| IDX_CUST | 1 | 34 | 2 (0)| 00:00:01 |
------------------------------------------------------------------------------
Predicate Information (identified by operation id):
PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------
---------------------------------------------------
1 - filter(:LNAME2>=:LNAME1)
2 - access("CITY"=:2 AND "LNAME">=:LNAME1 AND "LNAME"<=:LNAME2)
filter("LNAME">=:LNAME1 AND "LNAME"<=:LNAME2)
Note
-----
- dynamic statistics used: dynamic sampling (level=2)
20 rows selected.
这里满足第三颗星，但第一 第二颗星只能二选一
避免排序 选择第二颗星
最窄索引片 选择第一颗星
扫描二维码关注公众号，回复：
1675311 查看本文章
